Shadows not forming

Thu Sep 08, 2016 4:42 am

I think I am pushing this project to the limits of Octane, but not really sure. In short, this scene is probably 95% lit by indirect lighting through all of the pushbuttons. Sure, there are four big direct lights (the big red rectangle and another three out of camera view), that offer by far the majority of illumination power, but quantity-wise, the pushbuttons are the major burden. And possibly because of this, there is flaky behavior in the rendering of shadows (see image). They are either not getting completely filled in, or they are getting inverted in brightness. Also, all of the blue pushbuttons for example have the same power setting, but some render very quickly, while most others take forever to render.

Has anyone seen this behavior before and if so, does anyone know of a solution?

Beppe OctaneRender™ Italia Blue Sky · Thu Sep 08, 2016 11:06 am

Hi treddie,
difficult to say and very particular condition

since you have a lot of emissive materials with different power, did you already play with the sampling rate of any single emitter node in order to balance its importance in the general illumination?
ciao beppe

Thu Sep 08, 2016 1:15 pm

play with ray_epsilon

Thu Sep 08, 2016 1:16 pm

..something very familiar about that scene....

Everything that is going on appears to be in the nooks and crannies. Are you sure you don't have a dirt node plugged into this material(s), and it is causing it to be lighter in nooks? Seems like an easy fix though, with an AO pass popped on top in post.

Fri Sep 09, 2016 12:15 am

..something very familiar about that scene....

You must be as old as myself to recognize that interior!

since you have a lot of emissive materials with different power, did you already play with the sampling rate of any single emitter node in order to balance its importance in the general illumination?

Well, if I do that, won't those eventually be too bright when the noise clears up? One thing I tried was taking all the blue buttons and making them one object before importing into Octane. I did the same for the white and green buttons. I thought maybe then Octane would treat the blue group as one emitter with equal importance. Same for green and white. But the buttons closest to the display screens still solve much faster than the ones farther out.

Maybe a portal in front of each button.

play with ray_epsilon

I tried that but could not find a happy spot in the range. I'll try again.

...Are you sure you don't have a dirt node plugged into this material(s), and it is causing it to be lighter in nooks? Seems like an easy fix though, with an AO pass popped on top in post.

No dirt node anywhere. Exactly how does AO work? The problem as I see it is that since that is a post function, the under-exposed buttons will still be too noisy for AO to do much with them.

I've tried messing around with ray-epsilon and direct light importance, but no-go so far. Maybe Octane is just choking on so many emitters. AND indirect light emitters at that.

Tue Sep 13, 2016 2:45 am

Tried really hard with trial & error to find a sweet spot for ray-epsilon. No luck there. For reference, I have been using PMC exclusively. Am trying PT now.

Tue Sep 27, 2016 9:57 pm

Finally figured it out. It wasn't Ray-Epsilon. It was Direct Light Importance that was the problem. I lowered it all the way down, since there are not many direct light emitters visible to the camera, but tons of indirect lights. For this scene, dropping Direct Light Importance to its minimum not only cleared up the shadow issues, it spread the rate of light development evenly over the entire scene for all the push-buttons. It also sped the render up probably by a factor of 10 or more.

Beppe OctaneRender™ Italia Blue Sky · Wed Sep 28, 2016 8:29 am

Hi treddie,
great to hear that you have solved, and good point about Direct Light importance in PMC kernel. Congrats also for the final result, great atmosphere

ciao beppe

Wed Sep 28, 2016 6:53 pm

Thanks ciao beppe!

For those not familiar with the scene, this is a reverse-engineer of the space pod set in 1968's 2001: A Space Odyssey. It would only have taken one year to do it accurately, but a mistake I made early on forced me to have to revise hundreds of part files that took an extra year! Oh well.

I am hoping someone will build a full scale pod for a scifi museum somewhere, based on the engineering drawings I will be releasing in about another year. Now that the basic 3D solid model is complete, the next step is to make sure it can actually be built and assembled without issues. Standard fasteners and support brackets need to be added to the model, all of which will be hidden from view inside the mockup. And it has to be built so that eventual sagging will not occur for large thin parts. That is why it is spec'd out for 1/8" thick aluminum panels. It will be very heavy! But will also be on wheels so that it can be moved around. The spherical shell will have "wild" panels like the original set had.